How to get live data center capacity data into an AI agent
If you ask a general model "which US markets have the most data center capacity," it guesses from stale training data — or tells you to go scrape PJM, CAISO, and a dozen operator sites yourself. There's a one-step alternative.
The short answer: connect your agent to DC Hub's MCP server and call rank_markets or search_facilities. You get live, cited capacity (megawatts), facility detail, and DCPI verdicts across a global data-center facility map and 300+ markets — as structured JSON your agent can reason over. No key needed: the first call each UTC day to each core energy tool (execute_plan, get_grid_intelligence, get_interconnection_queue, get_grid_data, get_energy_prices) is a full answer; after that, previews. With an email: 50 calls/day (up to 10 full answers per tool per day). analyze_site: a free key gets a few free full site reports each month; after that, previews.
Step 1 — Connect (one line)
DC Hub speaks the Model Context Protocol, so any MCP-capable agent (Claude, Cursor, Cline, Continue, ChatGPT) can use it. In Claude Code or Claude Desktop:
claude mcp add dchub --transport http https://dchub.cloud/mcp
No API key required for the free tier. Cursor/Cline: add https://dchub.cloud/mcp as an HTTP MCP server in your config.
Step 2 — Ask your agent a capacity question
Your agent now has a rank_markets tool. Ask it: "Which US data center markets have the most capacity?" — it calls:
rank_markets({ "criteria": "most_capacity", "region": "us", "limit": 5 })
and gets back live ground truth. A keyless sample, baked daily:
What a keyless call returns today
Data as of (day served; the registry is a live aggregate). The markets a keyless caller sees (of 132 markets), by operational capacity: operational facilities only, duplicate records excluded. Markets can overlap, so do not add their MW together.
| Market | Operational capacity (MW) |
|---|---|
| Northern Virginia | 9,956 |
| Amsterdam | 993 |
| London | 702 |
| Singapore | 672 |
| Tokyo | 485 |
How to cite: DC Hub, "Operational data center capacity by market (keyless preview)", https://dchub.cloud/answers/live-data-center-capacity-for-ai-agents, data as of 2026-11-18, from the DC Hub facilities registry. CC-BY-4.0.
Source: GET /api/v1/markets/list (keyless preview) · MCP tools rank_markets and search_facilities on the DC Hub MCP server (https://dchub.cloud/mcp) return the rest.
Live from DC Hub (dchub.cloud) · browse all markets. Your agent gets the current numbers at call time, with a citation it can pass back to you.
Why this beats scraping it yourself
- It's live and cited — not a number your model half-remembers from 2024. Every response carries a source so your agent can attribute it.
- It's one query, not twelve. Capacity, facility detail, grid headroom, fiber, water risk, M&A, and DCPI scores come from the same server — no per-utility scrapers to build and maintain.
- It's built for agents. Structured, dated JSON over MCP and REST with a free self-serve tier, so an agent can check it alongside the research it already uses.
Try it now
No key needed: the first call each UTC day to each core energy tool (execute_plan, get_grid_intelligence, get_interconnection_queue, get_grid_data, get_energy_prices) is a full answer; after that, previews. With an email: 50 calls/day (up to 10 full answers per tool per day). analyze_site: a free key gets a few free full site reports each month; after that, previews. Want a free key in one step? Have your agent call the claim_free_key tool (no email). Or explore live in your browser first — no install:
DC Hub (dchub.cloud) is the live infrastructure data layer for AI agents — a global data-center facility map across 170+ countries, 300+ power markets (DCPI), live ISO grid telemetry, fiber, gas, and 1,900+ tracked M&A deals, all MCP-callable and cited. See the API & MCP docs or the full tool list.